Selective serotonin reuptake inhibitors and 5-HT₂ receptor agonists have distinct, sleep-state dependent effects on postictal breathing in amygdala kindled mice

Abstract

Seizures can cause profound breathing disruptions. Seizures arising from sleep cause greater breathing impairment than those emerging from wakefulness and more often result in sudden unexpected death in epilepsy (SUDEP). The neurotransmitter serotonin (5-HT) plays a major role in respiration and sleep-wake regulation. 5-HT modulates seizure susceptibility and severity and is dysregulated by seizures. Thus, the impact of seizures on breathing dysregulation may be due to impaired 5-HT neurotransmission. We examined whether pharmacologically increasing 5-HT neurotransmission prior to seizures improves postictal breathing and how sleep-state during seizure induction contributes to these effects. We assessed breathing with whole-body plethysmography in 84 amygdala-kindled mice pre-treated with selective serotonin reuptake inhibitors (SSRI) or 5-HT₂ receptor agonists. SSRIs and 5-HT₂ agonists increased postictal breathing frequency (f_R), tidal volume (V_T), and minute ventilation (V_E) at different timepoints following seizures induced during wakefulness. These effects were not observed following seizures induced during NREM sleep. SSRIs suppressed ictal and postictal apnea regardless of sleep state. The SSRI citalopram and the 5-HT₂ agonists TCB-2 and MK-212 decreased breathing variability following wake-occurring seizures at different postictal timepoints. Only MK-212 decreased breathing variability when seizures were induced during NREM sleep. The 5-HT_2A antagonist MDL-11939 reduced the effect of citalopram on f_R, V_T, and V_E, and enhanced its effect on breathing variability in the initial period following a seizure. These results suggest that 5-HT mechanisms that are dependent on or independent from the 5-HT₂ family of receptors impact breathing on different timescales during the recovery of eupnea, and that certain serotonergic treatments may be less effective at facilitating postictal breathing following seizures emerging from sleep.

INTRODUCTION:

Epilepsy is an extremely prevalent neurological disorder in which patients experience spontaneous recurrent seizures (Fisher et al., 2014). It is estimated that one in 26 Americans will develop epilepsy at some point during their lifetime (Hesdorffer et al., 2011; Kotsopoulos et al., 2002). Despite the availability of anti-seizure medications (ASM), approximately 35% of patients will not achieve seizure freedom with medical treatment (Chen et al., 2018; Kwan & Brodie, 2000). Seizures can cause profound alterations in breathing (Bayne & Simon, 1981; Kennedy et al., 2015; Rugg-Gunn et al., 2016). The degree of respiratory dysregulation often depends on seizure type and origin, with longer seizure duration being associated with a greater degree of respiratory dysfunction (Bateman et al., 2008; Bayne & Simon, 1981; Kennedy et al., 2015). Breathing dysfunction following seizures is exceptionally relevant to clinicians and individuals with epilepsy, as it is implicated as a primary mechanism underlying sudden unexpected death in epilepsy (SUDEP) (Buchanan et al., 2014; Kim et al., 2018; Ryvlin et al., 2013).

SUDEP refers to the non-traumatic, non-drowning death of a person with epilepsy, and is the leading cause of death among individuals with refractory epilepsy (Devinsky et al., 2016; Nashef et al., 2012). While the exact etiology of SUDEP is unknown, evidence suggests that it typically occurs following a generalized seizure (Nashef et al., 1998; Surges et al., 2009), and terminal apnea is the first step in this deadly cascade (Buchanan et al., 2014; Kim et al., 2018; Ryvlin et al., 2013). Timely mechanical ventilation or resuscitation greatly reduces seizure-induced mortality in both human patients and animal models (Buchanan et al., 2014; Ryvlin et al., 2013; Tupal & Faingold, 2006), further establishing the critical role respiratory disruption plays in SUDEP as well as the necessity of respiratory-focused interventions.

Apnea can occur both during and after seizures. Ictal apnea is relatively common, occurring in ~33–50% of focal seizures (Lacuey et al., 2018; Tio et al., 2020; Vilella et al., 2019). However, these apneas are typically brief and generally do not substantially impact arterial oxygen saturation (SaO₂) (Bateman et al., 2008). Ictal apnea has been documented in several animal models of epilepsy and SUDEP, including the Scn1a^R1407X/+ mouse model of Dravet syndrome—a childhood epileptic encephalopathy that has a high incidence of SUDEP (Dravet, 2000; Dravet et al., 2005; Kim et al., 2018). Seizure-induced respiratory arrest (S-IRA) in these animals can be prevented by mechanical ventilation (Kim et al., 2018). The DBA/1 and DBA/2 mouse models of audiogenic seizures also exhibit S-IRA that can be rescued via mechanical ventilation (Tupal & Faingold, 2006; Tupal et al., 2019). Postictal apnea is less common than ictal apnea (18% of GTCS) (Vilella et al., 2019). Postictal apnea occurs in both focal and generalized epilepsies, suggesting a separate pathophysiology from ictal apnea which occurs exclusively in focal epilepsy (Lacuey et al., 2018; Vilella et al., 2019). Postictal apnea is also associated with longer recovery times for hypoxemia and has been proposed as a potential SUDEP biomarker (Vilella et al., 2019). Patients with epilepsy (PWE) also exhibit a higher incidence of obstructive sleep apnea (OSA) (Popkirov et al., 2019; Lin et al., 2017). This may speak to a decrease in nighttime airway patency or a blunting of the hypercapnic arousal response in PWE (Joyal et al., 2022; Buchanan, 2019).

Respiratory rate variability is another potential respiratory biomarker of SUDEP. Previous investigations have revealed that mice that go on to die from a maximal electroshock (MES) seizure have increased respiratory rate variability at baseline compared with those that did not die (Hajek & Buchanan, 2016). The genetic Kcna1-null mouse model exhibits progressive respiratory dysfunction with age, including increased respiratory rate variability, along with spontaneous seizure-induced death (Dhaibar et al., 2019; Simeone et al., 2018). Further, human patients with increased respiratory variability have higher rates of postictal hypoxemia (Sainju, 2021). Hypoxemia itself may be a relevant biomarker for SUDEP. The duration of postictal hypoxemia is associated with potentially high-risk cardiac arrythmias (Park et al., 2017). Ictal oxygen desaturations are accompanied by increases in end-tidal carbon dioxide (ETCO₂), which suggests that this desaturation is a result of hypoventilation (Bateman et al., 2008).

Another major factor in augmenting SUDEP risk is vigilance state during seizure inception. Patients who die of SUDEP are twice as likely to have a history of nocturnal seizures (Lamberts et al., 2012; van der Lende et al., 2018). Breathing and cardiac activity are regulated in a sleep state-dependent manner (Buchanan, 2013; Joyal et al., 2022). Inspiratory drive is lower during non-rapid eye movement (NREM) sleep and lowest during rapid eye movement (REM) sleep (Douglas et al., 1982). Airway patency and the ventilatory response to hypercapnia are likewise reduced during sleep (Cherniack, 1981; Haxhiu et al., 1987). It follows then that nocturnal seizures are associated with more severe cardiorespiratory consequences, such as greater oxygen desaturation, compared to diurnal seizures (Hajek & Buchanan, 2016; Latreille et al., 2017). Previous work has found a decrease in breathing frequency (f_R), tidal volume (V_T), and minute ventilation (V_E) following MES seizures induced during NREM sleep compared to wakefulness (Hajek & Buchanan, 2016; Purnell et al., 2017). Patients are also more likely to be unaccompanied while asleep, and thus less likely to have someone available to administer potentially life-saving interventions (Rugg-Gunn et al., 2016; Sveinsson et al., 2020).

The neurotransmitter serotonin (5-HT) plays a major role in breathing and sleep-wake regulation and has been implicated in both epilepsy and SUDEP (Petrucci et al., 2020; Richerson, 2013; Richerson & Buchanan, 2011). Increasing 5-HT tone has been found to decrease seizure frequency in epilepsy patients (Ceulemans et al., 2012; Favale et al., 2003), while depletion of 5-HT or loss of 5-HT neurons increases seizure susceptibility in several rodent models of epilepsy (Buchanan et al., 2014; Kilian & Frey, 1973). 5-HT also plays a major role in breathing (Hodges et al., 2009; Richerson, 2004). Lower postictal serum 5-HT levels have been associated with postictal central apnea (Murugesan et al., 2019). Pre-treatment with 5-HT-augmenting agents can reduce and even block S-IRA following audiogenic and MES-induced seizures in mice without increasing basal respiration (Kruse et al., 2019; Buchanan et al., 2014; Tupal & Faingold, 2006; Zeng et al., 2015). Patients taking SSRIs experience reduced incidence of ictal hypoxemia (Bateman et al., 2010). Among 5-HT receptors, the 5-HT₂ family of receptors are known to play a critical role in the generation of respiratory rhythms (Hodges & Richerson, 2008b) as well as the modulation of respiratory motor neurons (Brandes et al., 2006).

Postictal respiration is a critical target for therapeutic intervention. Uncovering potential targets to improve postictal breathing is critical for the development of further therapies to mitigate SUDEP risk. In this study, we hypothesized that pharmacologically increasing 5-HT neurotransmission prior to a seizure would improve postictal respiration in an amygdala kindling model of epilepsy. As SUDEP commonly occurs at night (Lamberts et al., 2012; Purnell et al., 2018) and 5-HT is regulated in a sleep/wake-dependent manner (McGinty & Harper, 1976; Sakai, 2011), the effects of seizures induced during both wake and NREM sleep were assessed.

EXPERIMENTAL PROCEDURES:

Ethical Approval:

All procedures and experiments were approved by and performed in compliance with the Institutional Animal Care and Use Committee (IUCUC) at the University of Iowa Carver College of Medicine. All experiments complied with the National Research Council’s guide for the care and use of laboratory animals (National Research Council, 2011). Care was taken to minimize the number of animals used in this study as well as any possible pain and distress.

Experimental animals:

84 male and female C57BL/6J animals (17–29 g; Jackson Laboratories, Bar Harbor, ME) and 17 male and female 5-HT_2C KO mice (originally provided by Joel Elmquist, University of Texas Southwestern Medical Center, Dallas TX; Xu et al., 2008) were housed in a 12:12 light:dark cycle with food and water available ad libitum. Experiments were performed 2–6 hrs after the onset of the light phase. Experiments were spaced at least two days apart to reduce the confound of repeated stimulations. At the conclusion of all experimental trials, the animals were euthanized with an overdose of ketamine/xylazine (i.p., 50–75 mg/kg; 5–7.5 mg/kg).

EEG/EMG headmount and amygdala electrode implantation:

All surgeries were performed with aseptic technique under inhaled isoflurane anesthesia (1%–5% induction; 0.5%–2% maintenance). EEG/EMG headmounts (8201; Pinnacle Technology Inc., Lawrence, KS) were implanted as previously described (Buchanan & Richerson, 2010). Briefly, the skull was exposed and the headmount was attached to the skull via machine screws (1.0 mm thread × 4.1 mm length with 1.4 mm head diameter, #804052; Vigor Optical, Carlstadt, NJ) affixed to four holes (2 mm anterior to bregma/lambda; ±2 mm from midline). EMG wires emanating from the posterior of the headmount were inserted into the bilateral nuchal muscles ± 1–2 mm from the midline. Polyimide coated, bipolar stimulating/recording electrodes (MS333-3-BIU-SPC; Plastics One, Inc; Roanoke, VA) with the distal 0.5 mm of insulation removed were implanted into the right basolateral amygdala (in mm from bregma, AP: −1.3, ML: −2.8, DV: −4.7) as previously described (Petrucci et al., 2021). The stainless-steel ground wire was soldered to a ground screw placed between the left EEG electrodes. The headmount base, screw heads, amygdala electrode, and EMG wires were secured with dental cement (Jet Acrylic; Lang Dental, Wheeling, IL), and the skin was sutured closed, leaving only the headmount socket exposed. Animals received pre- and post-operative analgesia with meloxicam (2.0 mg/kg s.c. pre-operatively; 1.0 mg/kg/day s.c. post-operatively for 2 days) and were allowed to recover for at least 7 days before being studied.

EEG/EMG data acquisition:

EEG and EMG data were acquired as described previously (Buchanan et al., 2014; Buchanan & Richerson, 2010; Hajek & Buchanan, 2016). Briefly, animals were fit with a preamplifier (8202-SL; Pinnacle Technology) attached directly to the implanted headmount, introduced to the recording chamber, and allowed to acclimate as described below. Preamplifier leads were then passed through a commutator (8204; Pinnacle Technology) and into an analog conditioning amplifier (model 440 instrumentation amplifier; Brownlee Precision, San Jose, CA). EEG and EMG signals were amplified by 50,000 and bandpass filtered (0.3 to 200 Hz for EEG; 10 to 300 Hz for EMG). Data was digitized at 1,000 Hz with an analog-to-digital (A-D) converter (PCI-6221; National Instruments, Austin, TX) on a Dell desktop computer and acquired using software custom written in MATLAB (The MathWorks, Natick, MA).

Sleep-wake determination:

Sleep state was assessed online in real time before delivery of the amygdala stimulation. Vigilance state was assigned using a standard approach (Buchanan and Richerson 2010; Franken et al. 1998) based on EEG/EMG frequency characteristics as follows: Wake: low-amplitude, high-frequency (7–13 Hz) EEG with high EMG power; NREM: high-amplitude, low-frequency (0.5–4 Hz) EEG with moderate to low EMG power and lack of voluntary motor activity. Fast-Fourier transform power spectra were created with MATLAB for each 10-s epoch of data and used along with EEG and EMG characteristics to verify scoring post hoc.

Seizure induction:

Animals underwent amygdala kindling utilizing a rapid-kindling paradigm as described previously (Petrucci et al., 2021). An EEG/EMG preamplifier (8202-SL; Pinnacle Technology; Lawrence, KS) and a 3-channel electrode cable (335–340/3; Plastics One) were affixed to the animal’s headmount and connected to a data conditioning amplifier (LP511AC; AstroNova; West Warwick, RI), a stimulating/recording switch (SRS-13C0113G; AstroNova), and a pulse stimulator (Model 2100; A-M Systems). Amygdala afterdischarge threshold for each animal was determined by administering 20 μA incremental currents every 2 minutes until afterdischarges were recorded. The threshold stimulus was applied twice daily at least 1 hr apart (80–500 μA range, 1 s train of 1 msec biphasic square wave pulses at 60 Hz) until the animal was fully kindled. Seizures were objectively scored using a modified Racine scale. Kindling was considered complete following three consecutive Racine grade 4 (bilateral myoclonus, loss of posture, wild running) seizures (Buchanan et al., 2014; Racine & Coscina, 1979). Over the course of the kindling process, animals were acclimated to the recording apparatus.

Drugs:

After surgical recovery and completion of kindling, mice received an i.p. injection of citalopram hydrobromide (20 mg/kg), fluoxetine hydrochloride (10 mg/kg), TCB-2 (10 mg/kg), MK-212 hydrochloride (10 mg/kg), vabicaserin hydrochloride (30 mg/kg), or vehicle. To further delineate the mechanism by which citalopram modulates postictal breathing, two 5-HT₂ receptor antagonists were administered in conjunction with citalopram: including MDL-11939 (10 mg/kg), and RS-102221 (10 mg/kg). Vabicaserin was obtained from MedChemExpress. All other experimental drugs were obtained from Tocris Biosciences through Bio-Techne (Minneapolis, MN). Dosages were based upon the lab’s previous experience with the drugs and the primary literature (Buchanan et al., 2014). Drugs were diluted to the appropriate concentration with saline (0.9% NaCl). MDL-11939 was dissolved in 10% DMSO and then brought up to volume using corn oil (Sigma; St. Louis, MO). Vehicle treatments consisted of 150 μl injections of saline or 10% DMSO in corn oil.

Whole-body plethysmography:

To quantify respiration before and after seizure induction, a plethysmography recording chamber was fitted with a high-sensitivity/ultra-low-pressure transducer (DC002NDR5; Honeywell International, Minneapolis, MN). The analog output from the pressure transducer was digitized by an A-D converter (PCI-6221; National Instruments) displayed on a computer monitor in real time using an acquisition program custom written in MATLAB and saved on a computer hard drive. The breathing signal was calibrated prior to seizure induction by delivering metered breaths (300 μl; 150 breaths/min) to the recording chamber via a mechanical ventilator (Mini-Vent; Harvard Apparatus). Individual breaths were identified and measured using custom software written in MATLAB to aid in assessment of breathing parameters including breathing f_R, interbreath interval (IBI), V_T, and V_E as previously described (Buchanan et al., 2014; Hajek & Buchanan, 2016; Hodges & Richerson, 2008a). V_T was calculated using standard methods (Buchanan et al., 2014; Drorbaugh & Fenn, 1955) using relative humidity, ambient temperature, body temperature, and barometric pressure (https://www.wunderground.com). Apneas were manually scored by visualizing 10 s epochs of individual breaths in MATLAB. Respiratory data from 100 s prior to stimulation and 300 seconds after seizure termination were analyzed in 10-s epochs. f_R, IBI, V_T, and V_E were not scored during the seizure itself due to movement artifact. Epochs of pre- and postictal breathing with movement artifact, due to the animal sniffing or moving around the chamber, were also excluded from analysis.

Statistical analysis:

Estimates of the effects of epochs, drugs, and sleep state on normalized postictal f_R, V_T, and V_E were computed using multivariate linear mixed effect models, in which a random intercept was provided for each animal in the study. An estimate was included for the each of the effects being modeled, as well as an interaction term between the drug and epochs, to see whether the drug alters the effect of time on postictal f_R, V_T, or V_E. A series of Poincaré analyses were used to calculate the short-term standard deviation (SD1) and long-term standard deviation (SD2) of the IBI from the mean. For SD1 and SD2, a mixed effect Gamma regression model with a log link was used to account for the right-skewed nature of the data. The model construction procedure was otherwise the same as described above. Coefficients comparing f_R, V_T, V_E, SD1, and SD2 values between the wake and NREM conditions within each drug we reported from contrasts applied to mixed models including sleep state, drug, and their interaction (Table 1). Coefficients comparing f_R, V_T, V_E, SD1, and SD2 values between epochs within the NREM and WAKE conditions were reported from these models (Table 2).

Table 1:

Comparison of breathing frequency (f_R), tidal volume (V_T), minute ventilation (V_E), and short-term (SD1) and long-term (SD2) interbreath interval variability between wake vs NREM when animals received saline (SAL), citalopram (CIT), fluoxetine (FLX), and MK-212 (MK).

	SAL				CIT
	WAKE mean	NREM mean	Difference 95% CI	p-value	WAKE mean	NREM mean	Difference 95% CI	p-value
fR	0.713	1.130	0.102, 0.732	0.010	1.225	1.286	−0.485, 0.607	0.826
VT	1.279	2.206	0.067, 1.788	0.035	1.552	1.664	−1.329, 1.553	0.879
VE	1.03	1.71	0.111, 1.248	0.019	1.81	1.64	−1.143, 0.795	0.723
SD1	0.219	0.178	0.693, 0.956*	0.012	0.160	0.156	0.727, 1.303*	0.857
SD2	0.274	0.238	0.751, 1.002*	0.053	0.192	0.185	0.741, 1.248*	0.770
	FLX				MK
	WAKE mean	NREM mean	Difference 95% CI	p-value	WAKE mean	NREM mean	Difference 95% CI	p-value
fR	1.174	1.358	−0.36, 0.73	0.505	1.341	1.002	−0.917, 0.238	0.248
VT	1.549	1.990	−0.77, 1.66	0.476	2.883	1.982	−2.575, 0.772	0.290
VE	1.370	1.922	−0.32, 1.809	0.213	1.80	1.60	−1.267, 0.879	0.721
SD1	0.180	0.169	−0.41, 0.26*	0.676	0.145	0.132	0.652, 1.260*	0.558
SD2	0.211	0.201	−0.35, 1.25*	0.731	0.196	0.159	0.604, 1.086*	0.159

Confidence interval, CI.

^*Difference expressed as ratio of means rather than difference due to values being generated using a Gamma model with a log transformation.

Table 2:

Pooled comparisons of breathing frequency (f_R), tidal volume (V_T), minute ventilation (V_E), and short-term (SD1) and long-term (SD2) interbreath interval variability between the 0–100 s vs 101–200 s postictal epoch and 0–100 vs 201–300 s postictal epoch across all vehicle trials.

WAKE
	0–100 s Mean	101–200 s Mean	Difference 95% CI	p-value	0–100 s Mean	201–300 s Mean	Difference 95% CI	p-value
fR	1.815	0.350	−1.83, −1.13	<0.001	1.815	0.027	−2.13, −1.45	0.350
VT	3.621	1.173	−3.52, −1.45	<0.001	3.621	−0.611	−5.20, −3.17	1.173
VE	3.013	0.408	−3.23, −1.98	<0.001	3.013	−0.214	−3.84, −2.61	0.408
SD1	0.082	0.270	1.01, 1.30*	<0.001	0.082	0.270	1.19, 1.47*	0.270
SD2	0.148	0.307	0.55, 0.88*	<0.001	0.148	0.374	0.78, 1.10*	0.307
NREM
	0–100 s Mean	101–200 s Mean	Difference 95% CI	p-value	0–100 s Mean	201–300 s Mean	Difference 95% CI	p-value
fR	2.287	0.839	−1.74, −1.08	<0.001	2.287	0.222	−2.38, −1.75	<0.001
VT	4.138	2.054	−3.17, −0.86	<0.001	4.138	0.358	−4.88, −2.68	<0.001
VE	3.524	1.147	−2.91, −1.68	<0.001	3.524	0.315	−3.79, −2.63	<0.001
SD1	0.076	0.233	0.95, 1.24*	<0.001	0.076	0.258	1.10, 1.37*	<0.001
SD2	0.139	0.286	0.53, 0.88*	<0.001	0.139	0.320	0.69,1.01*	<0.001

Confidence interval, CI.

^*Difference expressed as ratio of means rather than difference due to values being generated using a Gamma model with a log transformation.

Data availability statement:

All raw data and analyses available upon request.

RESULTS

SSRIs and 5-HT₂ receptor agonists had distinct effects on postictal f_R, V_T, and V_E following seizures induced during wakefulness but not NREM sleep.

Of the C57BL/6J animals implanted, 2 animals were found dead-in-pen (DIP) with full hind-limb extension. This suggests they may have died from a spontaneous seizure. One animal died during its final kindling stimulation. To determine whether increasing endogenous 5-HT improves postictal breathing, the SSRIs citalopram and fluoxetine were systemically administered to animals prior to seizures induced during either wakefulness or NREM sleep. Breathing was assessed immediately following seizure termination until 300 s post-seizure. In order to assess how breathing changed over time as the animal recovered from a seizure, the postictal period was divided into three 100 s epochs. All breathing measures were compared to 100 s immediately prior to seizure induction.

Linear mixed effects modelling revealed a significant difference in mean postictal f_R, V_T, and V_E between wakefulness and NREM sleep (Table 1). There was also a significant difference in f_R, V_T, and V_E between the early postictal epoch (0–100 s) and the middle (101–200 s) and late (201–300 s) epochs when animals received vehicle treatment (Table 2; wake, n = 23; NREM, n = 24).

Citalopram (CIT, 20 mg/kg; wake: n = 13, NREM n = 12) increased f_R and V_E during the early and middle postictal epoch and had no effect on V_T when seizures were induced during wakefulness (Fig 1, Table S1). Fluoxetine (FLX, 10 mg/kg; wake, n = 11; NREM, n = 10) increased f_R and V_E during the middle postictal epoch and increased V_T during the middle and late postictal epochs following seizures induced during wakefulness (). There was no difference in postictal f_R, V_T, or V_E with SSRI treatment compared to saline after seizures induced during NREM sleep (Fig 1, Table S1). When a slightly higher dose of fluoxetine was administered there was no longer a difference in f_R, V_T, or V_E compared to vehicle during either sleep state (Fig S1; 20 mg/kg; wake, n = 8; NREM, n = 8). This is possibly due to a decrease in the specificity of fluoxetine binding to the serotonin transporter at higher concentrations (Stahl, 1998; van Harten, 1993).

Figure 1: Citalopram and fluoxetine increased ventilation during different postictal epochs following seizures induced during wakefulness but not NREM sleep.

Postictal (A) f_R, (B) V_T, and (C) V_E following seizures induced during wakefulness or NREM sleep in animals receiving saline (SAL, wake n = 23; NREM n = 24) or citalopram (CIT, 20 mg/kg; wake n = 13; NREM n = 12). (D) Average seizure duration following citalopram pretreatment. Postictal (E) f_R, (F) V_T, and (G) V_E following seizures induced during wakefulness or NREM sleep in animals receiving SAL or fluoxetine (FLX, 10 mg/kg; wake n = 11; NREM n = 10). (H) Average seizure duration following fluoxetine pretreatment. All values are relative to 100 s of preictal baseline breathing. *, p < 0.05; **, p < 0.01; ***, p < 0.001.

To assess whether stimulating 5-HT₂ receptors facilitates postictal breathing, 5-HT₂ receptor agonists MK-212 and TCB-2 were systemically administered to animals prior to seizures induced during either wakefulness or NREM sleep. The 5-HT_2C agonist MK-212 (MK, 10 mg/kg; wake, n = 9; NREM, n = 10) consistently increased postictal f_R, V_T, and V_E during the middle and late postictal epoch when seizures were induced during wakefulness (Fig 2, Table S2). There was no difference in postictal f_R, V_T, or V_E with MK-212 compared to saline after seizures induced during NREM sleep (Fig 2). The 5-HT_2A agonist TCB-2 (TCB,10 mg/kg; wake n = 11) decreased postictal f_R and V_E during the early postictal epoch in animals that had a seizure induced during wakefulness (Fig 1, Table S2). TCB-2 caused sleep to become extremely fragmented and transient. Thus, we were unable to induce seizures during NREM sleep after pretreatment with TCB-2.

Figure 2: MK-212 but not TCB-2 increased ventilation following seizures induced during wakefulness but not NREM sleep.

Postictal (A) f_R, (B) V_T, and (C) V_E following seizures induced during wakefulness or NREM sleep in animals receiving saline (SAL, n = 23; NREM n = 24) or MK-212 (MK, 10 mg/kg; wake n = 9; NREM n = 10). (D) Average seizure duration following MK-212 pretreatment. Postictal (E) f_R, (F) V_T, and (G) V_E following seizures induced during wakefulness in animals receiving SAL or TCB-2 (TCB, 10 mg/kg; wake n = 11). (H) Average seizure duration following TCB-2 pretreatment. All values are relative to 100 s of preictal baseline breathing. *, p < 0.05; **, p < 0.01; ***, p < 0.001.

As longer seizures are associated with a greater degree of respiratory dysfunction (Bateman et al., 2008; Bayne & Simon, 1981; Kennedy et al., 2015), we quantified the average duration of seizures in animals that received each drug treatment. There was no difference in seizure duration following administration of citalopram, fluoxetine, regardless of sleep state. TCB-2 slightly increased seizure duration compared to saline (SAL: 39.00 ± 3.816 vs TCB-2: 59.91 ± 5.667; p = 0.0057; Fig. 2D) and MK-212 slightly increased seizure duration when seizures were induced during NREM sleep (SAL: 30.00 ± 2.81 vs MK: 49.09 ± 3.59, p = 0.002; Fig 2D).

SSRIs but not 5-HT₂ agonists eliminated ictal and postictal apnea.

Ictal and postictal apnea are common occurrences in PWE (Lacuey et al., 2018; Tio et al., 2020; Vilella et al., 2019). While numerous animal models of epilepsy are used to assess apnea and S-IRA, apneas have not been characterized in an amygdala kindling model. In order to characterize the incidence of apnea in amygdala kindled animals and determine the effect of serotonergic treatments on their occurrence, we quantified the incidence and duration of apneas occurring during and after amygdala kindled seizures. For the purposes of this investigation, an apnea was defined as IBI ≥ 1 s.

Ictal apnea during amygdala kindled seizures was common, but not ubiquitous. Nearly 40% of animals that received saline and underwent seizure induction while awake experienced ictal apnea, while 30% of animals that were in NREM sleep during seizure induction experienced ictal apnea. This estimate might be low; however, as movement artifact during seizures likely obscured several apneas. Postictal apnea was considerably less common—animals only experienced postictal apnea when they underwent saline trials during NREM sleep or TCB-2 trials during wakefulness. Even then, the occurrence was just under 20% for both conditions. The majority of postictal apnea occurred immediately upon seizure termination, with animals rarely experiencing a postictal apnea more than 30 s post-seizure. For that reason, postictal apnea was scored for 100 s following seizure termination. When animals were pretreated with citalopram or fluoxetine (10 mg/kg) they exhibited no ictal or postictal apnea, regardless of sleep state during seizure induction (Fig 3A,C). This effect on postictal apneas persisted when the dose of fluoxetine was raised to 20 mg/kg (Fig S1). However, due to the relatively low occurrence of apnea in kindled animals, these findings did not reach significance.

Figure 3: SSRIs but not 5-HT₂ agonists eliminated apnea.

Representative traces of ictal apnea (A) and postictal apnea (B) emerging from saline trials. Top panel depicts the entire seizure. 10 s of EEG and plethysmography (pleth) data during the apneas. Temporal position of the traces are indicated by the red box in the above seizure trace. (C) Number of ictal apneas for all seizures induced during wakefulness (solid points) or NREM sleep (hollow points). (D) Duration of ictal apneas in animals that had seizures induced during wakefulness or NREM sleep. (E) Average number of postictal apneas per animal following seizures induced during wakefulness or NREM sleep. (F) Duration of postictal apnea in animals that had seizures induced during wakefulness or NREM sleep.

Despite its consistent effects on postictal f_R, V_T, and V_E, MK-212 pretreatment did not alter ictal apnea occurrence compared to saline during wakefulness or NREM sleep (Wake: p = 0.625; NREM p = 0.594) (Fig 3A,C). Animals pretreated with TCB-2 did not experience ictal apnea when seizures were induced during wakefulness; however, these animals did experience postictal apnea during wakefulness (Fig. 3A,C).

The average duration of ictal apnea was 1.575 ± 0.239 s and average duration of postictal apnea was 2.838 ± 1.15 s. However, incidences of protracted apnea lasting ~10 s occurred during saline NREM and MK-212 NREM trials (Fig 3B,D). This may further underscore the potential respiratory complications associated with seizures emerging from sleep. There were no significant differences in apnea duration between any of the conditions (Fig 3B,D).

Citalopram, MK-212, and TCB-2 decreased postictal breathing variability during different postictal epochs.

It has been reported that mice that go on to die from an MES-induced seizure exhibit increased respiratory rate variability compared with those that survive the seizure (Hajek & Buchanan, 2016). This highlights respiratory variability as a potential biomarker of SUDEP risk. Thus, we assessed respiratory variability compared to baseline from the animals that underwent the previously described seizure trials. Poincaré analysis was used to compare the variance of the inter-breath interval (IBI) between drug manipulations and sleep state. The average standard deviation from the mean was calculated for data points perpendicular (SD1) and parallel (SD2) to the line of identity for each Poincaré plot. SD1 is considered a measure of short-term variability and SD2 is considered a measure of long-term variability (Brennan et al., 2001).

Immediately following amygdala kindled seizures there was a period of extremely regular breathing, typically lasting between 10–30 s—roughly the duration of postictal generalized EEG suppression (PGES). When the animal’s EEG began to recover, their breathing became increasingly irregular (Fig 4). Thus, IBI variability tended to be at its lowest immediately after seizure termination and at its highest during the late postictal epoch.

Figure 4: Citalopram but not fluoxetine decreased postictal breath-to-breath variability differently when seizures were induced during wake versus NREM sleep.

Poincaré plots of the IBI of sequential breaths following pretreatment with (A) saline (SAL, n = 23; NREM n = 24), (B) citalopram (CIT, 20 mg/kg; wake n = 13; NREM n = 12), (C) fluoxetine (FLX, 10 mg/kg; wake n = 11; NREM n = 10), after seizures induced during either wakefulness (left column) or NREM sleep (right column). Data from 3 postictal epochs is plotted in each panel. Black line, line of identity (LOI). Mean SD1 (short-term variability) following seizures induced during wakefulness or NREM sleep following pretreatment with CIT (D) or FLX (E). Mean SD2 (long-term variability) following seizures induced during wakefulness or NREM sleep following pretreatment with CIT (F) or FLX (G). All values are relative to 100 s of preictal baseline breathing. Black line, line of identity (LOI). SD1 and SD2 example on left panel of (A) but applies to all Poincaré plots. n = 8–13. *, p < 0.05; **, p < 0.01; ***, p < .001.

Citalopram pretreatment decreased SD1 IBI variability during the early postictal epoch and decreased SD2 variability during the early and late postictal epochs when seizures were induced during wakefulness (Fig 4, Table S3). When seizures were induced during NREM sleep, there was a decrease in only SD2 during the late postictal epoch (). There was no difference in postictal IBI variability between pretreatment with fluoxetine and vehicle (Fig 4, Table S3).

MK-212 pretreatment consistently decreased SD1 and SD2 during the middle and late postictal epoch, regardless of sleep state during seizure induction (Fig 5, Table S4). TCB-2 also significantly decreased SD1 variability during the middle and late postictal epoch and SD2 during the late postictal epoch when seizures were induced during wakefulness (Fig 5, Table S4). Due to sleep fragmentation following TCB-2 treatment, it is unknown whether this effect also would have persisted when seizures were induced during NREM sleep. Notably, MK-212 and TCB-2 increased SD1 during the early postictal epoch when seizures were induced during wakefulness but not NREM sleep (Fig 5D).

Figure 5: 5-HT₂ agonists MK-212 and TCB-2 decreased postictal breath-to-breath variability differently when seizures were induced during wakefulness, but only MK-212 decreases breathing variability following seizures induced during NREM sleep.

Poincaré plots of the IBI of sequential breaths following pretreatment with (A) saline (SAL, wake n = 23; NREM n = 24), (B) MK-212 (MK, 10 mg/kg; wake n = 9; NREM n = 10), (C) TCB-2 (TCB, 10 mg/kg; wake n = 11), after seizures induced during either wakefulness (left column) or NREM sleep (right column). Data from 3 postictal epochs is plotted in each panel. Black line, line of identity (LOI). Mean SD1 (short-term variability) following seizures induced during wakefulness or NREM sleep following pretreatment with MK (D) or seizures induced during wakefulness following pretreatment with TCB (E). Mean SD2 (long-term variability) following seizures induced during wakefulness or NREM sleep following pretreatment with MK (F) or seizures induced during wakefulness following pretreatment with TCB (G). All values are relative to 100 s of preictal baseline breathing. SD1 and SD2 example on left panel of (A) but applies to all Poincaré plots. Black line, line of identity (LOI). n = 9–12. *, p < 0.05; **, p < 0.01; ***, p < .001.

5-HT_2C agonists increase f_R, V_T, and V_E during the late postictal epochs in WT animals and in the early postictal epochs in animals lacking 5-HT_2C receptors.

To further examine the necessity of the 5-HT_2C receptor on postictal respiration, we obtained a more selective 5-HT_2C agonist, vabicaserin. The above experiments were repeated in a separate cohort of C57BL/6J animals (n = 8) to measure postictal f_R, V_T, and V_E. Animals received pretreatment of vabicaserin (VAB, 30 mg/kg) and had seizures induced during wakefulness. Similar to MK-212, there was a significant increase in f_R compared to vehicle during the middle and late postictal epoch, and an increase in V_T and V_E compared to vehicle during the late postictal epoch (Fig 6, Table S5).

Figure 6: 5-HT_2C receptor activation was necessary for the effect of MK-212 and vabicaserin on f_R, V_T, and V_E during the late postictal period, but not earlier epochs.

Postictal f_R following pretreatment with (A) vabicaserin (VAB, 30 mg/kg) in C57BL/6J animals, (B) MK-212 (MK, 10 mg/kg) in 5-HT_2C KO animals, or (C) VAB in 5-HT_2C KO animals. V_T following pretreatment with (D) vabicaserin (VAB) in C57BL/6J animals, (E) MK-212 (MK) in 5-HT_2C KO animals, or (F) VAB in 5-HT_2C KO animals. V_E following pretreatment with (G) VAB in C57BL/6J animals, (H) MK-212 (MK) in 5-HT_2C KO animals, or (I) VAB in 5-HT_2C KO animals. All seizures induced during wakefulness. All values are relative to 100 s of preictal baseline breathing. n = 8. *, p < 0.05; **, p < 0.01.

Given the relatively poor specificity of drugs that target the 5-HT₂ family of receptors, we wanted to ensure that the effects of MK-212 and vabicaserin on postictal f_R, V_T, and V_E following seizures during wakefulness was due to activation of 5-HT_2C receptors and not an off-target effect. A total of 17 animals with a central deletion of 5-HT_2C receptors were implanted as previously described. Of the animals implanted, 3 were found DIP with full hind-limb extension, suggesting they may have died from a spontaneous seizure. 6 animals died during kindling. This increased incidence of death in 5-HT_2C KO mice is consistent with previous data (Applegate & Tecott, 1998; Tecott et al., 1995). The remaining 8 animals underwent the experiments described previously and received the same doses of either MK-212 or vabicaserin. All seizures were induced during wakefulness. MK-212 pretreatment in the KO animals significantly increased f_R during the early and middle postictal epochs. This is contrary to the results obtained from C57BL/6J animals, in which MK-212 pretreatment increased f_R and V_E during the middle and late but not early postictal epoch (Fig 6, Table S5). Vabicaserin pretreatment increased f_R, V_T, and V_E during the early postictal epoch in animals that lacked the 5-HT_2C receptor but had no effect on these measures during the middle and late epochs (Fig 6, Table S5).

MK-212 and vabicaserin decreased IBI variability in WT animals, but not in animals lacking the 5-HT_2C receptor.

Poincaré analysis of WT animals that received pretreatment with vabicaserin revealed identical results as MK-212—a consistent decrease in SD1 and SD2 during the middle and late postictal epochs, with an increase in SD1 during the early postictal epoch (Fig 7A,B, Table S6). When vabicaserin was given to 5-HT_2C KO mice prior to a seizure, this effect was lost; however, there was a decrease in SD2 during the early postictal epoch (Fig 7D,E, Table S6). Conversely, when MK-212 was administered to 5-HT_2C KO mice, there was no difference in SD1 compared to vehicle; however, there was a decrease in SD2 variability during all postictal epochs (Fig 7C,D, Table S6).

Figure 7: 5-HT₂ receptor activation was necessary for the effect of MK-212 and vabicaserin on long-term but not-short term breath-to-breath variability.

(A) Poincaré plots of the IBI of sequential breaths following pretreatment with following pretreatment with saline (SAL; left) or the 5-HT_2C receptor agonist vabicaserin (VAB, 30 mg/kg; right) prior to seizure induction in WT animals. (B) Poincaré plots of the IBI of sequential breaths following pretreatment with following pretreatment with SAL (left) or MK-212 (MK, 10 mg/kg; right) trials in 5-HT_2C KO animals. (C) Poincaré plots of the IBI of sequential breaths following pretreatment with following pretreatment with SAL (left) or VAB prior to seizure induction in 5-HT_2C KO animals. Analysis of mean postictal SD1 following pretreatment with (D) saline (SAL) or vabicaserin (VAB) in C57BL/6J animals, (E) SAL or MK-212 (MK) in 5-HT_2C KO animals, and (F) SAL or VAB in 5-HT_2C KO animals. Analysis of mean postictal SD2 following pretreatment with (G) SAL or VAB in C57BL/6J animals, (H) SAL or MK-212 (MK) in 5-HT_2C KO animals, and (I) SAL or VAB in 5-HT_2C KO animals. All seizures were induced during wakefulness. Data from 3 postictal epochs is plotted in each panel. Black line, line of identity (LOI). SD1 and SD2 example on left panel of (A) but applies to all Poincaré plots. All values are relative to 100 s of preictal baseline breathing. n = 8. *, p < 0.05; **, p < 0.01; ***.

5-HT_2A receptors modulated the effect of citalopram on postictal f_R, V_T, and V_E.

Thus far, SSRIs and 5-HT₂ receptor agonists appeared to exert their effects on postictal respiration during distinct time periods following seizure termination. In order to further probe the mechanism by which SSRIs affect postictal respiration and whether there is any contribution of 5-HT₂ receptors, we performed experiments using 5-HT₂ receptor antagonists in conjunction with citalopram. Citalopram was chosen over fluoxetine as fluoxetine did not affect postictal IBI variability.

A separate cohort of C57BL/6J animals (n = 8) underwent seizure trials where they received an i.p. injection of either the 5-HT_2A antagonist MDL-11939, the 5-HT_2C antagonist RS-102221, or vehicle 15 min prior to citalopram pretreatment. Saline served as the vehicle for RS-102221 while 10% DMSO was the vehicle for MDL-11939. As citalopram was only effective at increasing postictal breathing following wake seizure trails, all seizures were induced during wakefulness.

When MDL-11939 was administered prior to citalopram, there was a sharp decrease in f_R, V_T, and V_E during the early postictal epoch, but an increase in f_R during the middle epoch (Fig 8, Table S7. There was no significant difference in ventilation when RS-102221 was administered prior to citalopram (Fig S2, Table S7).

Figure 8: 5-HT_2A receptor blockade modulated the effect of citalopram on postictal f_R, V_T, and V_E.

Postictal (A) f_R, (B) V_T, and (C) V_E at 3 postictal timepoints following pretreatment with either vehicle (10% DMSO) or the 5-HT_2A receptor antagonist MDL-11939 (MDL, 10 mg/kg) prior to citalopram (CIT, 20 mg/kg) administration and seizure induction. All values are relative to 100 s of preictal baseline breathing. All seizures were induced during wakefulness. n = 8. *, p < 0.05; ***, p < .001.

5-HT_2A and 5-HT_2C receptors modulated the effect citalopram on postictal breath-to-breath variability and apnea.

In order to assess the role of 5-HT₂ receptors in citalopram’s effect on postictal IBI variability, Poincaré analysis was performed on the data from the MDL-11393 and RS-102221 trials. When animals received MDL-11939 prior to citalopram, they exhibited an increased SD1 IBI variability during the early postictal epoch and a reduced SD1 and SD2 IBI variability during the late postictal epoch compared to when they received vehicle prior to citalopram (Fig 9, Table S8). There was a significant difference in SD2 IBI variability during the early postictal epoch when animals received RS-102221 prior to citalopram compared to vehicle (Fig 9, Table S8).

Figure 9: 5-HT_2A and 5-HT_2C receptor blockade modulated the effect of citalopram on postictal breath-to-breath variability.

(A) Poincaré plots of the IBI of sequential breaths following pretreatment with following pretreatment with either vehicle (10% DMSO; left) or the 5-HT_2A receptor antagonist MDL-11939 (MDL, 10 mg/kg; right) prior to citalopram administration and seizure induction. (B) Poincaré plots of the IBI of sequential breaths following pretreatment with following pretreatment with either vehicle (SAL; left) or the 5-HT_2C receptor antagonist RS-102221 (RS, 10 mg/kg; right) prior to citalopram administration and seizure induction. Analysis of (C) mean SD1 (short-term variability) and (D) mean SD2 (long-term variability) for DMSO+CIT and MDL+CIT trials. Analysis of (E) mean SD1 and (F) mean SD2 for SAL+CIT and RS+CIT trials. All seizures were induced during wakefulness. All values are relative to 100 s of preictal baseline breathing. Data from 3 postictal epochs is plotted in each panel. SD1 and SD2 example on left panel of (A) but applies to all Poincaré plots. Black line, line of identity (LOI). n = 8. *, p < 0.05; **, p < 0.01; ***, p < .001.

To evaluate the role of 5-HT₂ receptors in citalopram’s effect on suppressing seizure-induced apnea, ictal and postictal apneas were scored. Animals that received MDL-11939 + citalopram, 10% DMSO + citalopram, and saline + citalopram experienced no ictal and postictal apnea. When animals that received RS-102221 + citalopram there was a reemergence of ictal apnea in 25% of the animals tested (2/8 animals), but no postictal apnea (data not shown).

DISCUSSION:

Severe dysregulation in breathing follows some generalized seizures, particularly nocturnal seizures (Bayne & Simon, 1981; Kennedy et al., 2015; Rugg-Gunn et al., 2016). As a modulator of breathing, sleep-wake regulation, and seizure threshold/severity, the 5-HT system is a salient target for mechanistic investigation and potential therapeutic intervention (Petrucci et al., 2020; Richerson, 2013; Richerson & Buchanan, 2011). Here we demonstrated the effect of several serotonergic agents on postictal respiration in an amygdala kindling model of epilepsy.

We provided evidence that postictal breathing is regulated in a sleep-state dependent manner. There was a larger increase in f_R, V_T, and V_E as well as a decrease in IBI variability when animals received vehicle compared to baseline following seizures induced NREM sleep compared to those induced during wakefulness (Table 1). The larger increase in postictal ventilation may reflect the lower inspiratory drive in humans during NREM sleep and thus a compensatory effect following seizures (Douglas et al., 1982). This follows as preictal f_R, V_T, and V_E were lower in animals that were in NREM sleep before having a seizure induced (Table 2). Our lab has previously demonstrated that MES-induced seizures in mice that are induced during NREM sleep are associated with greater respiratory dysfunction than those induced during wakefulness (Hajek & Buchanan, 2016). Conversely, there was no difference between wake and NREM sleep in any of the drug conditions (Table 1), and our results indicate that the serotonergic treatments at a dosage of 10–20 mg/kg were ineffective at increasing f_R, V_T, and V_E following seizures induced during NREM sleep (Fig 1,2). This may speak to the necessity of a certain baseline level of endogenous 5-HT in order for these serotonergic agents to affect respiration. As 5-HT tone is lower during NREM sleep (Sakai, 2011; Trulson & Jacobs, 1979), even 5-HT reuptake is inhibited, there may be too little endogenous 5-HT available to facilitate breathing.

Despite both being SSRIs, citalopram and fluoxetine had strikingly different effects on post-seizure breathing, with citalopram having much more consistent effects in augmenting postictal breathing. Also of note, when the dose of fluoxetine was increased, there was a counterintuitive decrease in its efficacy in increasing postictal ventilation and suppressing ictal apnea. This is possibly due to a decrease in the specificity of fluoxetine to the serotonin transporter as the concentration increases. Citalopram is considered a more 5-HT-selective SSRI (Stahl, 1998; van Harten, 1993). Meanwhile, fluoxetine has many off-target effects, including inhibiting reuptake of norepinephrine (NE) and binding affinity for the 5-HT_2C receptor, among others (Stahl, 1998; van Harten, 1993). Previous work demonstrates that inhibition of NE reuptake reduces S-IRA in both the MES and DBA/1 mouse models (Kruse et al., 2019; Zhang et al., 2017). This may suggest that selective targeting of the 5-HT system drives the efficacy of SSRI treatment in increasing post-seizure ventilation, while NE may play a larger role driving autoresuscitation, a protective cardiorespiratory reflex that can serve as a “last ditch” effort to prevent terminal apnea, which often takes the form of large, spontaneous gasps (Erickson, 2020). The role of NE in breathing is well documented and undoubtably plays a role in the mechanisms behind periictal respiratory dysfunction (Whelan & Young, 1953; Haxhiu et al., 2003). Future investigations delineating the role of 5-HT and NE in this pathway are certainly warranted.

There was significant change in f_R, V_T, V_E, and IBI variability over the postictal period. The middle and late epochs were significantly different than the early epoch for all breathing measures (Table 2). This underscores the dynamic nature of the postictal period. This is also reflected in cortical activity, which recovers from the ictal state in stages: with various frequency bands returning to baseline activities at distinct postictal timepoints (Petrucci et al., 2021). Not only do f_R, V_T, V_E, and IBI vary across the postictal period, pharmacologic manipulations that target the 5-HT system exert effects during specific postictal timepoints. Citalopram consistently increased postictal breathing during the early post-seizure epoch. Citalopram also increased f_R, V_E, and long-term IBI variability (SD2) during the middle and late postictal epoch, in addition to the early epoch. These variations in efficacy at discrete post-seizure timepoints could highlight separate receptor mechanisms that are differentially affected by seizures and subsequently operate on distinct timescales during recovery of eupnea.

The 5-HT_2C agonists MK-212 and vabicaserin consistently augmented postictal respiration during the middle and late postictal epoch, namely increasing postictal f_R, V_T, and V_E (Fig 2,6A) while decreasing short-term and long-term breath-to-breath variability (Fig 5,7A). 5-HT_2C receptors are expressed in many respiratory nuclei, including but not limited to the pre-Bötzinger complex (preBötC), the nucleus tractus solitarius (NTS), the retrotrapezoid nucleus (RTN), and the hypoglossal motor nucleus (Clemett et al, 2000; Hodges & Richerson, 2010). The preBötC is known for its role as a central pattern generator for inspiratory rhythm (Del Negro et al., 2018; Richter et al., 2003). Thus, activation of 5-HT_2C receptors in this region may drive the decreased variability in IBI seen in animals treated with MK-212 and vabicaserin. However, unilateral ablation of the preBötC has been shown to disrupt respiration during NREM sleep but not wakefulness (McKay & Feldman, 2008). Despite this, MK-212 decreased IBI variability when seizures were induced during both wakefulness and NREM sleep (Fig 4). It is important to consider that, while consciousness may be impaired, animals are not asleep following seizures that are induced during sleep. Additionally, 5-HT_2C receptors are also expressed on spinal motor neurons that are downstream from the preBötC (Murray et al., 2011). Thus, 5-HT_2C agonists may directly stimulate motor neurons that produce the respiratory rhythm.

Somewhat unexpectedly, 5-HT_2C agonists increased short-term breathing variability during the early postictal epoch in WT animals and increased f_R, V_T, and V_E in animals lacking central 5-HT_2C receptors during the same epoch. This may indicate an antagonistic effect of 5-HT_2C receptor activation on respiration immediately following seizure termination that evolves into facilitation of respiration during the later postictal epochs. This may be explained by the seemingly contradictory effects of 5-HT_2C receptors in the NTS, where 5-HT_2C receptor activation has been observed to both increase excitatory neurotransmission and inhibit synaptic activity (Austgen et al., 2012; Sévoz-Couche et al., 2000). It is therefore possible that 5-HT_2C receptor activation modulates NTS activity via presynaptic inhibition and postsynaptic excitation. Recruitment of these presynaptic receptors immediately after a seizure may explain the inhibition of several breathing measures, while postsynaptic 5-HT_2C receptor activation facilitates respiratory drive as breathing evolves in the later postictal period.

TCB-2 pretreatment exhibited inconsistent effects on postictal breathing. Notably, it failed to prevent postictal apnea. This is somewhat contradictory to previous work where TCB-2 but not MK-212 eliminated S-IRA following MES seizures (Buchanan et al., 2014). This is possibly due to the difference in epilepsy models utilized in these studies. The MES model features generalized seizures that originate in the brainstem, whereas kindled seizures are focal to bilateral tonic clonic seizures originating from the limbic system (Kandratavicius et al., 2014). A further study found that optogenetic activation of 5-HT_2A receptors in the DBA/1 model of audiogenic seizures also prevented S-IRA (Shen et al., 2020). It is possible that 5-HT_2A receptor activation plays a larger role in facilitating autoresuscitation than other facets of postictal respiration. Gasping activity both in vitro (Tryba et al., 2006) and in vivo (Cummings et al., 2009), is dependent on 5-HT_2A activity. Blockade of 5-HT_2A receptors with MDL-11939 prior to citalopram significantly blunted the increase in f_R, V_T, and V_E during the immediate postictal period (Fig 8A). This suggests that 5-HT_2A receptor activation is necessary for the effect of citalopram on these postictal breathing measures, and thus may be necessary but not sufficient to increase postictal respiration.

Apnea is well-characterized in PWE, but up to this point they have not been characterized in an amygdala kindling model of epilepsy. Both ictal and postictal apnea were observed during and after amygdala kindled seizures during both wakefulness and sleep. SSRIs eliminated both ictal and postictal apnea, while the 5-HT₂ agonists did not, suggesting that there may be other 5-HT receptors involved in suppressing these apneas. 5-HT₂ receptors on hypoglossal motor neurons play a role in facilitating airway patency and dilation (Ogasa et al., 2004), which may be protective against obstructive apnea, but not central apnea. When 5-HT_2C receptors were blocked with RS-102221 prior to citalopram pretreatment, ictal apnea reemerged. This suggests an involvement of 5-HT_2C receptors in suppressing ictal apnea that is not sufficient to produce this effect alone. Despite citalopram losing its effect on overall ventilation and breathing stability when seizures were induced during NREM sleep, both citalopram and fluoxetine suppressed ictal and postictal apnea regardless of vigilance state. This should not be understated, as although ventilation and breathing stability are important for overall oxygenation, it is ultimately terminal apnea that is believed to result in SUDEP (Buchanan et al., 2014; Ryvlin et al., 2013). Due to a lack of direct measurements of airway patency, we are currently unable to determine whether these apneas were central or obstructive in nature.

A major limitation of our experimental design was the inability to extrapolate the effects of our manipulations to seizure-induced death, due to the limited incidence death in the amygdala kindling model. The kindling model was chosen due to the ability to induce seizures at will and due to the high survival rate, multiple trials could be performed in the same animals, yielding a robust, within-subjects comparison. However, the benefits of the increased respiration seen in our experiments cannot be linked to increasing seizure survivability unless shifted to a model that features seizure-induced death as a phenotype. Additionally, breathing could not be scored during seizures themselves due to movement artifact. Thus, the effects of our manipulation on breathing during the seizures themselves could not be measured with whole body plethysmography. While apneas were scored during seizures, it is likely that some were missed due to this artifact, thus our estimate of the percentage of animals that experienced ictal apnea is likely low. Additionally, the use of an acute, single dose of all of the drugs used in our experiments does not capture the types of changes in respiration that would come with a chronic dose that would be more typical of a treatment regimen PWE would be prescribed. Lastly, while we were not able to collect data during REM sleep during this particular study, the effects of REM sleep and the 5-HT system on post-seizure breathing is still a subject worth investigating.

Seizure-induced respiratory dysfunction is a major contributor to SUDEP pathophysiology (Buchanan et al., 2014; Rugg-Gunn et al., 2016; Ryvlin et al., 2013). In order to identify patients who may be at highest risk for SUDEP and develop therapeutic interventions, it is critical that we understand the mechanisms by which sleep impacts both postictal breathing and the effect of medications meant to improve said breathing. Our results suggest that treatments that solely target the 5-HT system may not be effective in increasing ventilation and breathing stability when seizures emerge from NREM sleep. However, SSRIs may suppress periictal apneas during both wakefulness and NREM sleep. We have also provided evidence that breathing evolves over time following seizure termination, and that different serotonergic receptor mechanisms modulate different facets of breathing on different postictal timescales. Future experiments will be necessary to explore further subtypes of 5-HT receptors in post-seizure breathing, including the 5-HT_1A, 5-HT_2B, 5-HT₃, and 5-HT₇ receptors. It is our hope that by untangling the role of 5-HT receptor mechanisms and sleep state in postictal respiration, we can offer clinicians specific pharmacological interventions that will be effective in protecting breathing following seizures at any time.

Highlights:

• SSRIs and 5-HT₂ agonists had differential effects on postictal breathing in kindled mice

• Different 5-HT receptor pathways modulate postictal breathing on distinct timescales

• Treatments were less effective at promoting breathing when seizures occurred during sleep

• Citalopram and fluoxetine pretreatment eliminated ictal and postictal apnea

Abbreviations:

5-HT

5-hydroxytryptamine, serotonin

AP

anterior-posterior

BLA

basolateral amygdala

DV

dorsal-ventral

CI

confidence interval

CIT

citalopram

EEG

electroencephalogram

EMG

electromyogram

ETCO₂

end-tidal carbon dioxide

FLX

fluoxetine

f_R

breathing frequency

IBI

interbreath interval

i.p.

intraperitoneally

KO

knockout

LOI

line of identity

MDL

MDL-11939

MES

maximal electroshock

ML

medial-lateral

MK

MK-212

NREM

non-rapid eye movement

Pleth

plethysmography

REM

rapid eye movement

RS

RS-102221

s.c.

subcutaneously

SD1

standard deviation of Poincaré values perpendicular to the line of identity

SD2

standard deviation of Poincaré values parallel to the line of identity

S-IRA

seizure-induced respiratory arrest

SaO₂

arterial oxygen saturation

SSRI

selective serotonin reuptake inhibitor

SUDEP

sudden unexpected death in epilepsy

TCB

TCB-2

VAB

vabicaserin

V_E

minute ventilation

V_T

tidal volume